AI Needs Power: 5 Dividend Stocks Quietly Funding the Datacenter Boom

The article argues AI-driven electricity demand is creating a strong setup for five utilities, with DOE projecting data centers could reach 12% of U.S. power demand by 2028. It highlights positive earnings beats, raised or affirmed EPS guidance, large regulated capex plans, and dividend growth at PSEG, Duke, Southern, Dominion, and NextEra. The most notable catalysts include PJM capacity prices rising to $329-$333/MW-Day, Duke's $103B capex plan, Dominion's board buying, and NextEra's 25-year Google nuclear PPA plus 9.5 GW gas buildout mandate.

Analysis

The market is still treating this as a yield trade, but the real shift is that regulated utilities are becoming the financing layer for AI infrastructure. That changes the earnings sensitivity: growth is no longer just load-driven, it is capex-driven, which tends to lock in multiyear rate base expansion and compress the probability of outright demand destruction. The second-order winner set is broader than the five names here: turbine OEMs, transformers, switchgear, gas pipeline interconnectors, EPC contractors, and even uranium supply should all see tighter order books as utilities and hyperscalers fight over delivery slots.

The most important dynamic is not who has the most load today, but who can translate load into allowed returns fastest. NEE and D have the cleanest near-term monetization because they already have project-specific counterparties and visible backlog; PEG is the most mispriced because nuclear optionality is under-credited and the stock is behaving like a defensive bond proxy despite a materially improving capacity-market backdrop. DUK and SO are more about capital formation than headline upside, but their scale means they can turn incremental large-load requests into decade-long earnings visibility.

The key risk is timing mismatch: demand can be marketed as immediate while revenue recognition depends on interconnection, permitting, fuel procurement, and rate-case approvals that can slip by quarters or years. If AI capex pauses, the sector can derate quickly because multiples already embed a smoother growth path than utility execution usually delivers. The contrarian view is that the market may be underestimating how much of this upside is already socialized into rate base and consensus, especially for DUK and SO; the cleaner mispricing may be the laggards with direct but less obvious nuclear/merchant leverage, not the obvious AI beneficiaries.

Near term, the trade is about sentiment persistence over the next 1-3 quarters; structurally, the setup can last 2-4 years if capacity shortages remain binding. The stock-specific tell will be whether management teams keep raising capex without equity dilution or cost-of-capital pressure. If that stays true, the current pullbacks should be bought rather than faded.

Key Decisions for Investors

• Long PEG on weakness over the next 1-2 weeks; target a re-rating toward 19-20x forward earnings if PJM pricing and nuclear utilization remain firm. Risk: if the market reverts to treating PEG as a pure utility bond proxy, upside compresses to dividend yield expansion only.
• Buy NEE for a 6-12 month move; use the recent drawdown as entry. The risk/reward is attractive because project-backed growth is still underappreciated, but a delay in the Duane Arnold timeline or gas buildout execution could cap near-term upside.
• Pair trade: long D / short XLU for 3-6 months. The thesis is that Dominion has the most direct Data Center Alley exposure and stronger insider signaling, while the ETF dilutes that optionality into slower names. Stop if utility multiples broaden across the sector on falling rates.
• Long DUK versus SO as a relative-value trade over 6-9 months. DUK’s capex scale and regulated growth visibility appear more underwritten than Georgia’s already-popular load story. Risk: SO could continue to outperform if commercial/wholesale growth keeps surprising.
• Avoid shorting the group outright; instead express skepticism via put spreads on the highest-multiple name in the basket if rates back up. The asymmetry is that upside can compound through rate base, while downside is mainly execution delay rather than demand collapse.